The Agent Skills Directory

[COMMAND_EXECUTION]: The skill allows the execution of arbitrary shell commands through the spawn configuration field in scenario files (e.g., ['node', 'dist/cli.js', '--yolo']). This is the core functionality required to generate terminal screenshots.
[DYNAMIC_EXECUTION]: Scenario configurations are stored in TypeScript (.ts) files which are dynamically loaded and executed by the run.ts entry point using npx tsx.
[INDIRECT_PROMPT_INJECTION]: The tool captures terminal output into screenshots for PR reviews. This creates a surface where malicious or confusing output generated by the spawned process could attempt to influence an agent that subsequently reviews the visual documentation.
Ingestion points: terminal output stream from node-pty rendered via xterm.js.
Boundary markers: None present in terminal capture output.
Capability inventory: Shell command execution (spawn), file system write (screenshots), and local script execution.
Sanitization: No sanitization of terminal output is mentioned before screenshot capture.
[EXTERNAL_DOWNLOADS]: The skill instructions facilitate the download of the Chromium browser via Microsoft's official Playwright distribution and the ffmpeg utility via the Homebrew package manager for GIF generation.

terminal-capture